Unintended pregnancies related to poor contraceptive adherence account for approximately 20% of the 3.5 million unintended pregnancies that occur in the United States annually.1 Patient nonadherence plays a large role in oral contraception failure with “forgetfulness” being the reason for 30% of nonadherence among patients.2,3 Investigators are turning to short message services, also known as text messaging, to improve patient adherence with daily medication regimens for treating a variety of medical conditions.4,5 However, the effect of text messaging on daily adherence with oral contraceptive pills has not been studied in a randomized controlled fashion.
The rate of true oral contraceptive pill nonadherence is underestimated in available contraceptive studies. Contraceptive trials traditionally use patient diaries as a standard for collecting adherence data.6 Because this conventional method overreports the consistency with which patients take medications, adherence studies are now expected to use more objective measurements.7,8 An electronic monitoring device with an embedded memory chip has been tested among oral contraceptive pill users, but data collection was limited by unreturned devices or erased memory chips.7 Technology has been developed since to collect adherence data in real time by wireless communication, which likely provides a more accurate picture of oral contraceptive pill adherence.
In this randomized controlled trial, our primary aim was to compare the oral contraceptive pill adherence of women who received daily text-message reminders with women who did not receive reminders. Because the patient pill diary does not reflect true oral contraceptive pill adherence, this study measured the number of missed pills using an electronic monitoring device with wireless real-time data collection. The secondary aim was to compare this device with traditional patient pill diaries as a tool for measuring adherence among all our participants.
MATERIALS AND METHODS
We conducted an investigator-blinded randomized controlled trial examining the effect of daily text-message reminders on oral contraceptive pill adherence. This study was approved by the Partners Healthcare Institutional Review Board. Study participants were recruited from a single Planned Parenthood League of Massachusetts clinic located in Boston, MA, an affiliate of Planned Parenthood Federation of America, from August 2008 through February 2009. Women seeking care at Planned Parenthood League of Massachusetts who were sexually active, had voluntarily chosen the pill as their contraceptive method, and were new users of the method were asked if they wished to participate in the study. Inclusion criteria included a personal mobile phone with text-messaging capabilities and desire to use oral contraceptive pills for at least 6 months. Exclusion criteria included contraindications to oral contraceptive pill use, the use of oral contraceptive pills primarily for noncontraceptive indications, current use of oral contraceptive pills for more than 1 month, intent of pregnancy in the subsequent 12 months, plans to travel outside the cell phone networks of the United States and Canada during the duration of the study, inability to complete the study, and being in situations in which receipt of a text-message reminder to take an oral contraceptive pill would endanger the safety and privacy of the woman.
After providing written informed consent for study participation, women completed baseline questionnaires and returned after a 1-month run-in period. Women were allowed to start oral contraceptives during this run-in period according to the standard clinic protocol. Women who returned for the second visit were assigned randomly, with equal probability to text-message reminders or no reminders, and given the electronic monitoring device and three cycles of oral contraceptive pills. We required a 1-month run-in period to increase the likelihood that the electronic monitoring devices loaned to participants for the study would be returned. Enrollment continued until the number of participants randomly assigned met the desired sample size. Randomization was computer-generated with varying blocks of four and six. Randomization sequences were concealed within opaque, sequentially numbered envelopes until interventions were assigned by a research assistant who did not participate in data analysis. Half of the participants received daily text-message reminders and the other half did not. During the 3-month study period, each participant assigned the intervention received a daily text message, “Please remember to take your birth control pill,” sent at a designated time chosen by the participant (TXT Signal, Inc., Gainesville, FL). Women in the control group did not receive reminders. Investigators involved in data analysis were blinded to group assignment until primary data analysis was completed.
The oral contraceptive pills dispensed in this study were combined monophasic hormonal pills in commercially available formulations that contain 35 mcg of ethinyl estradiol or less and various progestins packaged in a four-pill by seven-pill rectangular matrix. These included pill formulations with 21 active and seven placebo tablets or 24 active and four placebo tablets to be used in a cyclic fashion. We included both oral contraceptive pill formulations to retain study generalizability. In addition, we felt that allowing women some control over their pill choice may minimize the likelihood that they would stop using oral contraceptive pills altogether during the study and thus minimize the rate of attrition. The pill packs were modified so that each pack could be inserted into the electronic monitoring device, and each pack was placed in a separate envelope and labeled for the participant to place into the device at the beginning of its cycle. Three months of oral contraceptive pills were provided at no cost to study participants for the duration of the study. Each woman received an electronic monitoring device called SIMPill (Clinical Technology Advisors, Inc., Acton, MA) for adherence data collection. Women were told that the device would monitor their pill-taking during the study. Each time the participant opened the electronic monitoring device to remove a pill, a short message service signal was sent wirelessly to a main server and a second server. The main server recorded the date and time the electronic monitoring device was opened, which was contained within the short message service signal itself. The second server (The Center for Connected Health, Boston, MA) recorded when it received the short message service signal in the event that the data contained within the short message service signal was inaccurate. As proof of pill-taking, each day was examined for at least one record of electronic monitoring device opening on either server. All participants were asked to recharge the electronic monitoring device once a month and received a text message on approximately the 15th of every month of the study reminding them to do so.
In addition to using the electronic monitoring device, the women kept a diary of their daily pill-taking. Women received three monthly diary cards on which to record this information. Careful instructions were given to facilitate consistent and correct completion of the diary cards. At the end of 3 months, participants returned to the clinic for a third visit to complete a final questionnaire and to return their devices and pill diaries.
To differentiate between true poor adherence and systems malfunction, we formulated general rules for the adherence data. Each day was examined for at least one electronic monitoring device opening on either server. If seven or more electronic monitoring device openings in a cycle were recorded, we assumed that the device was functioning for the cycle. If four to six openings in a cycle were recorded and the patient assessed herself as nonadherent (missing three or more pills a month) on the final questionnaire, then we assumed the device was functioning. However, if four to six openings were recorded in a cycle but the patient assessed herself as adherent (missing two or fewer pills per month), we assumed that the electronic monitoring device was malfunctioning for the cycle and the cycle was dropped from analysis. If three openings or less were recorded in a given cycle, we assumed that the electronic monitoring device was malfunctioning and the cycle was dropped from analysis.
Like in other oral contraceptive pill studies using electronic monitoring devices, the operational definition of a pill-taking day was from 3:00 am to 2:59 am.7,9 Although a full pill cycle consists of 28 days, active hormonal pills are taken only during the first 21 or 24 days; inactive “reminder” pills (placebos) are taken for the remaining days. Only events during the first 21 active pill days of a cycle were used in the analysis.
The sample size calculation was based on a study showing that women missed an average of 2.6 pills per cycle as measured by an electronic monitoring device.7 To detect a 1.6-pill improvement with standard deviation of two pills, using a two-sided test at 90% power (α=0.05), we required a sample size of 34 patients per group. Anticipating a 15% dropout rate, we randomly assigned 82 women to ensure that 68 would complete the study with useable data.
Data were analyzed according to intention-to-treat principles with investigators blind with respect to arm assignment until the final report. The data were initially plotted and inspected for distributional assumptions and the possibility of erroneous outliers. The arms were compared for baseline characteristics using the chi-square or Wilcoxon rank-sum test. Missed-pill data were summarized using means and standard deviations. The overall analysis framework for electronic monitoring device data was a mixed-models analysis of variance with one between-participants main effect of treatment arm (text compared with control) and two within-participants main effects of cycle (one through three) and week (1 through 3) and participant as a random effect. The three-way interaction of treatment arm, cycle, and week was eliminated (P=.98) followed by the elimination of the interaction between cycle and week (P=.99). The criterion for rejecting the null hypothesis for each main effect was 5% α. The α level was adjusted for multiple comparisons with 2.5% for effects within each arm and 1.67% for effects within each cycle. The comparison of the electronic monitoring device and diary data was derived from an analysis similar to the model described previously with data-collection method as an additional within-participant factor. Analyses were performed using SAS 9.1 (SAS Institute, Inc., Cary, NC).
A total of 103 women enrolled in the study, and 82 returned for randomization after the 1-month run-in period. The demographics and reproductive histories of the 21 women who did not return after the 1-month run-in period differed significantly from those who returned for randomization. Those who did not return were less likely to have completed high school, less likely to be nulliparous, and more likely to have been recruited for the study while seeking an abortion (P<.05). The baseline characteristics of the women who were assigned randomly to text-message reminders did not differ significantly from those of the participants in the control group (results not shown). On average, women who were assigned randomly in the study were 22 years old and most had completed high school (99%). Most of the study participants were white (79%), nulliparous (99%), and considered themselves to be in a relationship (57%). More than half of the participants had been on oral contraceptive pills in the past (51 of 82 [62%]). Of these previous oral contraceptive pill users, 25 (49%) had used a reminder system with their previous pill use with 21 women using a cell phone or clock alarms. Condom and emergency contraception awareness was high in our population with 61 (74%) estimating condom use during at least 50% of coital acts in the last 5 years and 44 (54%) reporting emergency contraceptive use at least once in the past 5 years. Before randomization, 74 women (97% of responders) felt that they could rely on text-message reminders to help them remember to take their pill at least “sometimes.”
Of the 82 women randomized, 73 (89%) had usable electronic monitoring device data at the end of the study (Fig. 1). The other nine did not have any usable data as a result of a nonfunctioning device (n=1), a presumed nonfunctioning device (n=5), protocol violations (n=2), or the participant stopping her oral contraceptive pills within the first cycle (n=1). The demographics of these women did not differ significantly from the women who had data at the end of the study (data not shown). Of the 73 women with usable data, 48 had data for all three cycles. Fifteen women had data for two cycles as a result of a presumed nonfunctioning device for one cycle (n=11), dropping from the study (n=3), or stopping oral contraceptive pills during the third cycle (n=1). The final 10 women had data for only one cycle either because of a presumed nonfunctioning device for two cycles (n=8) or because she stopped her pills during the second cycle (n=2). The demographics and reproductive histories of the women with useable data did not differ significantly between the text message and control groups (Table 1).
The rate of missed pills for the text-message group and the control group as recorded by the electronic monitoring devices did not differ significantly: 4.9±3.0 missed pills per cycle for the text-message group and 4.6±3.5 missed pills per cycle for the control group (P=.60). The lack of difference between groups remained if analyses were restricted to women with data for all three cycles: 4.1±2.6 missed pills per cycle in the text-message group and 3.7±3.4 missed pills per cycle in the control group (P=.67). The number of missed pills increased with each cycle over the study for the entire cohort (P=.02) but did not increase differentially according to arm (P=.58) (Table 2; Fig. 2).
Throughout the study, there was a statistically significant increase in missed pills from week 1 to week 3 during cycles (P=.02). As shown in Figure 2, the text group started each cycle with slightly but not significantly more missed pills in the first week (P=.30). As the weeks progressed within the cycles, missed pills by women in the text-message group did not change very much (P=.71), whereas the increase in missed pills was striking among participants in the control group (P=.007).
More than one third of women in the text-message arm used a reminding system in addition to their text messaging (13 of 36 [36%]), whereas 68% of those in the control group (26 of 37) used a reminding system outside the study protocol (P=.003). The majority of these women used an alarm clock or a cell phone alarm to provide reminders (23 of 26 [88%]). Within the control group, the mean number of missed pills per cycle did not differ between women who used a reminding system and those who did not (P=.80)
All women felt that it was very important to remember to take their birth control pills with a median score of 10 on a scale from 0 (“not important at all”) to 10 (the most important thing she can do to prevent pregnancy.”) Despite the frequent use of reminding systems outside the study protocol, 47 women in the study (57%) missed an average of three pills or more per cycle, distributed evenly between the groups (P=.37). Overall, 19 women reported side effects with oral contraceptive pill use on their final questionnaire, equally distributed between the text-message group (six of 31 [19%]) and the control group (13 of 35 [37%]) (P=.11). In our study, although text messaging did not improve oral contraceptive pill adherence, 12 participants (16%) demonstrated excellent oral contraceptive pill adherence (missing an average of one or fewer pills per cycle) unrelated to intervention (P=.22). There were no obvious differences in demographic or reproductive characteristics between this subset of oral contraceptive pill users and the rest of the cohort, except for a higher rate of baseline reported condom use in the first group with at least 50% of coital activity (P=.04). However, significantly more of these oral contraceptive pill users with excellent adherence felt at the beginning of the study that text messaging would help them remember their pills only “a little” or “not at all” (P=.009).
Despite the number of pills missed in the study, there were no pregnancies reported during the study. The majority of women (68 of 74 [91%]) expressed pregnancy anxiety, agreeing or strongly agreeing with the statement “I would be very upset if I were to get pregnant right now.” Overall, 24 women (35%) reported condom use for at least 50% of coital activity during the study. Condom use during the study significantly differed between groups with 17 women (47%) in the text-message group and nine women (24%) in the control group reporting condom use for at least 50% of coital activity (P=.04). Only two participants reported emergency contraception use during the study, both in the text-message group.
In general, the 35 women who received the text-message reminders felt the reminder was useful with the median score of 8 on a scale of 0 (“It was useless”) to 10 (“I would never remember without it”). The majority of them relied on the text-message reminders with 66% stating that the text messages reminded them to take their pills more than 50% of the time. Many of the women receiving text-message reminders (30 of 35 [86%]) would continue or would consider continuing to use this reminder system, and 34 women (97%) would recommend or consider recommending it to a relative or friend. More than half of the women who received text-message reminders (20 of 35 [57%]) would pay to use the system with the median acceptable cost of less than $5 a month. Women also felt that pages, cell phone alarms, and providing more information about the problems associated with missing oral contraceptive pills would help remind them to take their pills.
We also compared the electronic monitoring device data to pill diary data among our participants (Table 2; Fig. 3). We received diary data for 61 women (74%), 58 submitting diaries for all three cycles. The overall rate of missed pills was 4.7±3.2 per cycle according to the electronic monitoring device and 1.2±1.5 per cycle according to patient diary (P<.001). Although the electronic data showed a mean of 4.9±3.0 pills missed in the text-message group, women in this group recorded a mean of 1.4±1.9 missed pills per cycle in their diaries (P<.001). Similarly, women in the control group, for whom electronic data showed a mean of 4.6±3.5 missed pills per cycle, reported 1.1±1.2 missed pills per cycle (P<.001).
The majority of women completing the final survey (57 of 74 [77%]) stated that they preferred the wireless device over the pill diary for collecting adherence data because the device required less patient effort with 30 (41%) noting difficulties with filling out the diary as a result of forgetfulness. The remainder (17 of 74 [23%]) preferred the pill diary, attributing this preference to the electronic monitoring device's inconvenient size and mechanism.
Although oral contraceptive pills are an effective form of contraception if used correctly, many women are at risk for unplanned pregnancy as a result of missed pills. This study assessed the effect of daily text-message reminders on oral contraceptive pill adherence using electronic monitoring. Although women receiving text-message reminders felt that the text messages were useful and relied on them, there were no statistically significant differences in the mean number of missed pills per cycle between women who received daily text-message reminders and women who did not.
The lack of a significant difference between the arms could be attributed to the frequent use of alternative reminder systems in the control arm. At the beginning of the study, all women were given routine contraceptive counseling, including the possibility of using reminder methods to help them to remember to take their pill. The high rate of reminder system use in the control arm, particularly electronic systems such as cell phone alarms that mimicked the study intervention, could contribute to the lack of difference between the arms. Whether this use was the result of the extensive formal education or reproductive health knowledge of our study cohort, the counseling received, or study reactivity (in which participants may change their behavior simply in response to the fact that they are being studied) is unclear.
Despite the use of alternative reminding systems, the rate of missed pills in both arms was clinically high in the study population. Participants were randomly assigned after a run-in period meant to increase device return at the end of the study and all participants were aware that pill-taking would be tracked, so this low adherence in our study may still overestimate the adherence of the general population. Despite poor adherence, no pregnancies were reported during our study, although we might have expected one pregnancy based on the 8% risk of pregnancy during the first year of typical oral contraceptive pill use. Possible reasons for the lack of pregnancies may be that our participants had a low frequency of sexual activity or correctly used backup methods like condoms when they had poor adherence with oral contraceptive pills. Coital activity and condom use with each sexual act was not assessed. The text-message arm also reported more frequent condom and emergency contraception use during the study, so the text-message reminders may have had an unintended effect of reinforcing use of backup methods for women at risk.
Our study cohort was mostly white, well-educated, and nulliparous, which may limit the generalizability of these findings to a more diverse population. The women who did not return after the run-in period were less likely to have completed high school, more likely to have given birth, and more likely to have been recruited for the study while seeking an abortion. Whether these women would be less adherent with their oral contraceptive pills at baseline is unknown, and whether they might have benefited from text-message reminders is similarly unclear.
Our demonstrated mean rate of missed pills of 4.7 pills per cycle was higher than the rate of 2.6 mean pills missed reported by Potter et al, 7 although the average missed pills per cycle similarly increased over time. Because we cannot exclude the possibility that participants may open and close the electronic monitoring device without taking their oral contraceptive pill, the electronic monitoring device may be underestimating the actual missed pill rate in the cohort. However, the wireless data collection is likely a more accurate reflection of missed pill rates than the memory chips used in the Potter study, because the technology is not dependent on women returning the device at the end of the study. Our wireless technology was limited by mechanical and technological issues as described, but only 42 of the possible 246 cycles (17%) could not be included in analysis compared with 220 of a possible 507 cycles (43%) that could not be analyzed in the study by Potter et al. The increase in missed pills over time likely reflects a return to normal pill adherence behavior as study reactivity wanes. The significant increase of missed pills per week over the course of a cycle among the control participants suggests that beginning each new cycle of oral contraceptive pills may have enforced study reactivity that quickly diminished over the course of each cycle and that this effect was blunted by daily text-message reminders in the intervention group.
Our finding of a poorer adherence rate as measured by the electronic monitoring device compared with that of the patient diary is similar to the behavior observed by Potter et al.7 This supports the idea that although objective measures are more expensive, they provide a more accurate measurement of true adherence than self-report.8 Current study participants also agreed with this conclusion and preferred the electronic monitoring device over pill diaries because the device required less participant effort.
Previous studies have shown improvement in health outcomes with communication technologies, but a direct effect on medication adherence is unclear. A review of the role of cell phones and text messaging in improving health outcomes and processes of care identified 20 randomized controlled trials examining these interventions, but only one of these measured medicine taking adherence as an outcome.10 This study examined 58 adults with HIV taking highly active antiretroviral therapy. Half of the participants were provided with a programmable portable electronic device that gave verbal reminders and the other half received standard counseling.11 Adherence as measured by the electronic monitoring device was not significantly different among patients with normal cognition, but post hoc analysis of memory-impaired participants revealed significantly higher adherence rates among those receiving reminders. These findings among people with normal cognition are consistent with our results.
In summary, text-message reminders are insufficient to improve daily oral contraceptive pill adherence. Other strategies to improve adherence and continuation of hormonal contraception such as group motivation, structured, peer or multicomponent counseling, and intensive reminders of appointments have previously demonstrated no benefit for participants.12 In our study, although text messaging did not improve oral contraceptive pill adherence, a subset of women demonstrated excellent oral contraceptive pill adherence unrelated to intervention. This suggests that there are some women who are excellent pill takers and are good candidates for oral contraceptive pill use and others who will have poor pill adherence even with reminder systems. Additional studies are needed to help differentiate between those likely to be successful using oral contraceptive pills and those who would be better served using “forgettable contraception” to reduce the risk of unintended pregnancy.13
1. Rosenberg MJ, Waugh MS, Long S. Unintended pregnancies and use, misuse and discontinuation of oral contraceptives. J Reprod Med 1995;40:355–60.
2. Rosenberg M, Waugh MS. Causes and consequences of oral contraceptive noncompliance. Am J Obstet Gynecol 1999;180:276–9.
3. Cramer JA. Identifying and improving compliance patterns. In: Cramer JA, Spilker B, editors. Patient compliance in medical practice and clinical trials. New York (NY): Raven Press; 1991. p. 387–92.
4. Neville R, Greene A, McLeod J, Tracy A, Surie J. Mobile phone text messaging can help young people manage asthma [published erratum appears in BMJ 2008;336(7649)]. BMJ 2002;325:600.
5. Franklin VL, Waller A, Pagliari C, Greene SA. A randomized controlled trial of Sweet Talk, a text-messaging system to support young people with diabetes. Diabet Med 2006;23:1332–8.
6. Osterberg L, Blaschke T. Adherence to medication. N Engl J Med 2005;353:487–97.
7. Potter L, Oakley D, de Leon-Wong E, Canamar R. Measuring compliance among oral contraceptive users. Fam Plann Perspect 1996;28:154–8.
8. Haynes RB, Ackloo E, Sahota N, McDonald HP, Yao X. Interventions for enhancing medication adherence. The Cochrane Database of Systematic Reviews 2008, Issue 2. Art. No.: CD000011. DOI: 10.1002/14651858.CD000011.pub3.
10. Krishna S, Boren SA, Balas EA. Healthcare via cell phones: a systematic review. Telemed J E Health 2009;15:231–40.
11. Andrade AS, McGruder HF, Wu AW, Celano SA, Skolasky RL Jr, Selnes OA, et al. A programmable prompting device improves adherence to highly active antiretroviral therapy in HIV-infected subjects with memory impairment. Clin Infect Dis 2005;41:875–82.
12. Halpern V, Grimes DA, Lopez L, Gallo MF. Strategies to improve adherence and acceptability of hormonal methods for contraception. The Cochrane Database of Systematic Reviews 2006, Issue 1. Art. No.: CD004317. DOI: 10.1002/14651858.CD004317.pub2.
13. Grimes DA. Forgettable contraception. Contraception 2009;80:497–9.